KldsSeeGhosts

[KldsSeeGhosts @KldsSeeGh0sts]

@maria_rcks Am i tweaking or is deepseek harness not on here

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux No shot we’re skipping business plans… AGAIN

[KldsSeeGhosts @KldsSeeGh0sts]

@dannyveigatx @MaziyarPanahi There’s a plugin that works very well called dsh-tui

[KldsSeeGhosts @KldsSeeGh0sts]

@blueemi99 Genuinely speaking I was in the camp that it’s mimo because xiaomi could definitely do 100t. But I’m convinced it’s glm now solely because ZCode is actually working for once and not “always busy” they may have come across compute

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux Anthropic might shit in your mouth, but at the least they're pretty transparent when they do change around the usage

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux All due respect, this is a cop out and you know it Tibo, back months ago when I had a 1 month free trial of GPT plus it was near impossible to burn through the limits. Its okay to nerf things, its not okay to not be transparent about it.

[KldsSeeGhosts @KldsSeeGh0sts]

@Da7_Tech What are your thoughts on the devin subscription? I bought it for a month to use the free glm 5.2, and am thinking of getting another as they are extending glm 5.2 another month

[KldsSeeGhosts @KldsSeeGh0sts]

@bijanbowen Honestly probably a psycho idea but what if you ran the same prompt on each of the 3 or maybe even just glm / qwen in the video just to showcase what we have locally vs sota

[KldsSeeGhosts @KldsSeeGh0sts]

@Howaboua @pidotdev In regards to GPT Live, I've been meaning to work on a local clone / version with STT / TTS models running off my 5090, do you think this is something that could be done? It would be really cool to be able to use cloned voices etc.

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux Not gonna lie, biggest thing is I think (my opinion) the codex cli looks hideous. It’s genuinely vile to look at compared to Claude Code. The codex desktop app on the other hand is fucking beautiful

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux No Arch :/

[KldsSeeGhosts @KldsSeeGh0sts]

@Alibaba_Qwen @bijanbowen Are we going to see that qwen 3.8 video now that its out? I wasn't personally very impressed with the preview personally.

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux Please please please please allow for customization of more ui elements of Codex Cli, I think its one of the few weakpoints of it. Please allow more statusline customization like how Claude Code allows for designs with chevrons and icons.

[KldsSeeGhosts @KldsSeeGh0sts]

@bijanbowen @kimmonismus If they could somehow provide Luna for this plan or something in that ballpark ish I’d be all over it

[KldsSeeGhosts @KldsSeeGh0sts]

@thsottiaux @xikhar @supabase I just want you to know you have cost me many a 4-5am sleepless night. I want off this wild ride

[KldsSeeGhosts @KldsSeeGh0sts]

@lawrencecchen Super exciting, I’ve been looking over the fence for so long at cmux. Herdr is really nice and has been the best I’ve used on Linux so far, excited to see what the cmux team brings in v2! Good luck with the continued development!

[KldsSeeGhosts @KldsSeeGh0sts]

@theo Really need to get Pi implemented, it’s the only coding agent I use and I’d really like to try out t3 code as I’ve never liked desktop apps over clis before

[KldsSeeGhosts @KldsSeeGh0sts]

It’s absolutely absurd to me just how fucking useless google Ai models are jfc

[KldsSeeGhosts @KldsSeeGh0sts]

@cline Doesn't appear to be the case for me, says the promotion has ended when trying to use glm-5.2 (free)